Salesforce's 23% Rally: Repricing a Panic, or an EPS Illusion?

Salesforce just did something it hasn't done in more than two decades. After the enterprise-software company reported its fiscal-second-quarter results in late August, the shares jumped about 23% in a single session — their best month since May 2005 — after spending most of 2026 as one of the most hated stocks in the market. The stock had slid to a 52-week low near $146 during the year's AI selloff and was still down more than 20% for the year the day before the report. Now it sits near $243, up about 26% over the last month.

The title throws two options at a holder: take profits, or buy more. That framing deserves a sharper answer than "it depends." Because what drove the rally was a genuine repricing of a panic — wrapped around an earnings number that was partly an illusion.

For much of 2026, Salesforce was the poster child for the "SaaSpocalypse," the fear that cheap, capable AI agents would replace the seat-based software that built the entire software-as-a-service industry. The narrative erased roughly $2 trillion in market value from software stocks early in the year, and Salesforce fell about 40% from its highs. When the market punishes a business it thinks is about to be disrupted, the question is whether the revenue data actually shows it. For Salesforce, it didn't.

The late-August report didn't merely beat a set of low expectations — it showed the doom forecast was not materializing. Revenue rose 11% to $11.35 billion. More importantly, the forward-looking order book — current remaining performance obligations, the contracted sales expected to be recognized over the next year — grew 14% to $33.5 billion. AI-related revenue from Agentforce and Data 360 climbed more than 210% year over year to nearly $3.9 billion. Customer attrition hovered near its lowest level ever, and growth in net new order value was the strongest in four years. Those are the numbers that matter for a thesis that was supposed to be dying.

Salesforce also answered the "AI will replace us" fear by partnering with the presumed replacement. It expanded its relationship with Anthropic, maker of the Claude model, to build "Claudeforce," weaving Anthropic's AI into its software and its customer data. That one move reframed the whole story: instead of model makers disrupting software vendors, they were cutting deals with them.

Now the part the rally headlines minimized. The report showed adjusted earnings per share nearly doubling to $5.90 — a figure that helped fuel the surge. But roughly $2.53 of that per-share amount was investment gains, primarily the mark-to-market appreciation of Salesforce's own stake in Anthropic. That is not money earned by selling software; it is paper profit on a stock holding. Strip it out, and operating adjusted EPS was closer to $3.37, against the roughly $3.28 analysts had modeled. A real beat, but a modest one — not the doubling the headline implied.

That distinction reframes what a buyer at today's price is actually paying for. At roughly $243, Salesforce trades around 29 times forward earnings for a business growing revenue about 11%. The bargain that existed at $146 — where the market had arguably baked in the doomsday scenario — is gone. The buyer at the current price is no longer buying a panic discount; they are buying a bet that the AI order inflection compounds for several quarters to come. That is a legitimate bull case, especially given that aggregate rating services still label the stock a Buy and several banks raised their targets after the report. But it is a different risk than the one at the low.

So: take profits, or buy more? The honest answer is neither, reflexively. The rally corrected a real error — the market had priced a scenario the data did not support, and the repricing was rational. That is a point in favor of holders and of anyone who bought the panic. But the easy risk/reward has been collected. At 29 times forward earnings with roughly 11% top-line growth, the remaining upside rests on the AI numbers continuing to beat, and one quarter that flattered EPS with investment gains does not prove that. Morgan Stanley kept a neutral rating, saying it needed lasting evidence that the new AI products accelerate sales.

For a new buyer, the disciplined posture is the opposite of February's: don't chase a stock in the month after its best day in two decades. The re-rating was earned; the compounding has to be proven. Stripping the Anthropic gain out of the headline EPS shows how much of this move rests on narrative and how much on operating reality — and for fresh money at the top of a one-month spike, that gap is the whole risk.